Biom ELECTRA Large SQuAD2 BioASQ8B

🚀 BioM-Transformers: Building Large Biomedical Language Models with BERT, ALBERT and ELECTRA

This project empirically studies biomedical domain adaptation with large transformer models using different design choices and achieves state - of - the - art results on several biomedical domain tasks.

🚀 Quick Start

The README provides details about the BioM - Transformers model, including its fine - tuning process, usage, performance evaluation, training script, and citation information. You can use the provided training script to train the model and evaluate its performance on relevant datasets.

✨ Features

• The model is fine - tuned on SQuAD2.0 and BioASQ8B - Factoid training datasets.
• It can be used for direct prediction (inference) without further fine - tuning.
• Useful for creating a pandemic QA system (e.g., COVID - 19).
• Achieves state - of - the - art results on several biomedical domain tasks with similar or less computational cost.

📦 Installation

The README does not provide specific installation steps. You may need to refer to the original code repository for installation guidance.

💻 Usage Examples

Basic Usage

You can use the model to make predictions directly. Try entering a PubMed abstract in the context box in the model card and ask a couple of biomedical questions within the given context.

Advanced Usage

If you want to train the model, you can use the following training script:

python3 run_squad.py --model_type electra --model_name_or_path sultan/BioM-ELECTRA-Large-SQuAD2 \
--train_file BioASQ8B/train.json \
--predict_file BioASQ8B/dev.json \
--do_lower_case \
--do_train \
--do_eval \
--threads 20 \
--version_2_with_negative \
--num_train_epochs 3 \
--learning_rate 5e-5 \
--max_seq_length 512 \
--doc_stride 128 \
--per_gpu_train_batch_size 8 \
--gradient_accumulation_steps 2  \
--per_gpu_eval_batch_size 128   \
--logging_steps 50 \
--save_steps 5000 \
--fp16 \
--fp16_opt_level O1 \
--overwrite_output_dir \
--output_dir BioM-ELECTRA-Large-SQuAD-BioASQ \
--overwrite_cache

📚 Documentation

Model Description

• The model is fine - tuned on the SQuAD2.0 dataset and then on the BioASQ8B - Factoid training dataset. The BioASQ8B - Factoid training dataset is converted to SQuAD1.1 format for training and evaluation.
• It can be used for direct prediction without fine - tuning and is useful for creating a pandemic QA system.
• Note that the PyTorch version is different from the one used in the participation in BioASQ9B (TensorFlow with Layer - Wise Decay). All five batches of the BioASQ8B testing dataset are combined as one dev.json file.

Performance Evaluation

Below are the unofficial results of the models against the original ELECTRA base and large:

📄 License

The README does not provide license information.

📄 Citation

@inproceedings{alrowili-shanker-2021-biom,
title = "{B}io{M}-Transformers: Building Large Biomedical Language Models with {BERT}, {ALBERT} and {ELECTRA}",
author = "Alrowili, Sultan and
Shanker, Vijay",
booktitle = "Proceedings of the 20th Workshop on Biomedical Language Processing",
month = jun,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2021.bionlp-1.24",
pages = "221--227",
abstract = "The impact of design choices on the performance of biomedical language models recently has been a subject for investigation. In this paper, we empirically study biomedical domain adaptation with large transformer models using different design choices. We evaluate the performance of our pretrained models against other existing biomedical language models in the literature. Our results show that we achieve state-of-the-art results on several biomedical domain tasks despite using similar or less computational cost compared to other models in the literature. Our findings highlight the significant effect of design choices on improving the performance of biomedical language models.",
}

Acknowledgment

We would like to acknowledge the support we have from Tensorflow Research Cloud (TFRC) team to grant us access to TPUv3 units.